10-08-2018 12:47 AM - edited 10-08-2018 12:51 AM
Hi,Please give me a professional suggestion. Thank you very much.
I need to reinstall the operating system on 10 hosts, all with datanode, nodemanager, regionserver roles, and 10T HDFS data. The cloudera-scm-agent role is mounted on the system disk, and the HDFS data is on the special data disk. The cloudera manager version is 5.7.2, and the CDH version is 5.7.6.
The ideal operation is to reinstall two hosts at a time and redeploy coudera-scm-agent, nodemanager, regionserver, and datenode roles on the host, but reuse the HDFS data of the previous data disk, so that the normal external service of the cluster can be guaranteed and the HDFS triple replica recovery mechanism will not be triggered to ensure the cluster IO is normal.
But how do I implement my operation through cloudera manager?
The key points are as follows:
1. do I need to decommissioning the datanode, regionserver and nodemanager roles on the host?
2. how can I avoid the HDFS three replication recovery mechanism to reduce the cluster IO?
I hope I can get some more detailed suggestions on how to operate. Thank you！